JP2007276508A - Collision avoidance controller for vehicle - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain optimal manipulated variables of an actuator for making a vehicle travel on an optimal route where a collision is avoidable by model prediction control. <P>SOLUTION: This collision avoidance controller is provided with: a database (6) for storing data relating to the traveling state of its own vehicle and data relating to a distance between a target and its own vehicle and the relative speed of the target; a target future prediction part (2) for predicting the future location of the target based on the data stored in the database (6); its own vehicle future prediction part (3) for predicting the future location of its own vehicle based on the data stored in the database (6); a collision risk degree evaluation part (4) for evaluating the risk degree of a collision based on outputs of the future prediction parts of the target and its own vehicle; and an optimal manipulated variable retrieval part (5) for operating model prediction control based on the output of the collision risk degree evaluation part, and for retrieving optimal manipulated variables relating to the actuator of its own vehicle for avoiding the collision. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a vehicle collision avoidance control apparatus, and more particularly to a vehicle collision avoidance control apparatus using model predictive control.

  Various collisions that detect obstacles around the vehicle (hereinafter referred to as a target), determine the risk of collision with the target, and automatically set an avoidance operation on the vehicle when necessary An avoidance device has been developed. For example, in the collision prevention apparatus described in Patent Document 1, the possibility of collision is determined from the traveling state of the target traveling in front of the host vehicle and the traveling state of the host vehicle, and the collision should be avoided based on the determination result. The brake or accelerator is controlled. In this case, the collision avoidance operation is performed while the steering direction of the vehicle is fixed. Similarly, the apparatus described in Patent Document 2 determines the possibility of collision with the target, and when the possibility of collision is high, the brake is forcibly driven to avoid the collision.

  On the other hand, in the vehicle collision prevention apparatus described in Patent Document 3, when the vehicle is about to change lanes, if there is a vehicle with a high possibility of collision in the changed lane, the reaction force in the direction to prevent lane change is applied to the steering wheel. By trying to avoid collisions.

JP 7-104062 A JP2002-120679 JP 9-221052

  As described above, the conventional vehicle collision avoidance device controls only the traveling speed of the own vehicle or forcibly controls only the steering wheel when the possibility of a collision with the target is high. Collision avoidance to the immediate target is attempted by either operation of blocking lane change.

  However, even in order to avoid a collision in a running vehicle, suddenly applying a brake has a significant effect on other vehicles in the vicinity of the vehicle other than the target. There is a possibility of collision. In addition, forcibly preventing lane changes may increase other risks such as inability to avoid targets to be avoided. For this reason, an apparatus for controlling a brake or steering as disclosed in Patent Documents 1 to 3 to avoid a collision is not suitable for use on an actual road where a plurality of vehicles are running side by side. There was a problem in its practicality.

  The present invention was made for the purpose of avoiding the above-described drawbacks of the conventional vehicle collision avoidance device. When the risk of collision between the own vehicle and the target is detected, the own vehicle By calculating the optimum route to be taken and driving the brakes and steering, and in some cases, the accelerator, based on the calculation result, it is possible to smoothly avoid collisions without sudden changes in the running state. It is an object to provide a collision avoidance control device.

  In order to solve the above-described problems, a vehicle collision avoidance control apparatus according to the present invention includes a database unit that stores data relating to a traveling state of a host vehicle, data relating to a distance between the target and the host vehicle, and a relative speed of the target; A target future prediction unit that predicts the future position of the target based on the data stored in the unit, a vehicle future prediction unit that predicts the future position of the vehicle based on the data stored in the database unit, and the target And a collision risk evaluation unit that evaluates the risk of collision between the vehicle and the target based on the output of the future prediction unit of the vehicle, and model prediction control based on the output of the collision risk evaluation unit to avoid collision And an optimal operation amount search unit that searches for an optimal operation amount related to the actuator of the vehicle.

  According to the control device of the present invention, the host vehicle model and the target model are created from the data related to the traveling state of the host vehicle, the data about the distance between the target and the host vehicle and the relative speed of the target, and the model prediction control is performed. It is possible to search for the optimum operation amount related to the actuator for the vehicle to travel on the optimum route for avoiding the collision. By inputting this optimum operation amount to an actuator of the own vehicle, for example, a brake, a steering, and further an accelerator, the own vehicle can automatically travel on the optimum route to avoid a collision.

  Further, the optimum operation amount can be displayed on a display of a navigation device or the like and used for driving assistance of the driver. In addition, since the optimal operation amount is detected by model predictive control, even when there are multiple targets, search for the optimal route in consideration of multiple targets by creating a model for each target. Is possible. Therefore, it can fully cope with use on an actual road where a plurality of vehicles run side by side.

  FIG. 1 is a block diagram showing a schematic configuration of a vehicle collision avoidance control apparatus according to an embodiment of the present invention. In the figure, reference numeral 1 denotes a collision avoidance control device using model predictive control. By modeling the movement of the host vehicle and the target (obstacle), the optimal route for avoiding the collision between the two is searched, and It is a device for outputting the optimum operation amount of the actuator for the vehicle to travel on this route.

  The collision avoidance control device 1 includes a target future prediction unit 2 that predicts the future position of the target, a vehicle future prediction unit 3 that predicts the future position of the host vehicle, and the predicted future position of the target and the future position of the host vehicle. Collision risk evaluation unit 4 that evaluates the risk of collision between the vehicle and the target. If there is a risk of collision, create the vehicle model and the target model, and avoid collision based on the optimization algorithm. An optimum operation amount search unit 5 for retrieving the optimum operation amount of the own vehicle, and a database unit 6 that stores various data relating to the traveling state of the own vehicle and the traveling state of the target in time series.

  The optimum operation amount search unit 5 searches for an optimum operation amount to be input to the actuator 7 of the own vehicle in order to avoid a collision. Specifically, this operation amount is an operation amount related to the brake, steering, and accelerator of the own vehicle. In the optimum operation amount search unit 5, a temporary operation amount is input to the own vehicle model to calculate the travel route, and the collision risk is evaluated with respect to the calculated future position of the own vehicle and the future position of the target. The input value of the manipulated variable is optimized so that the degree of risk is the lowest, that is, the evaluation value is the smallest possible. The input value of the operation amount optimized in this way is output as an optimal operation amount to the actuator 7 of the own vehicle or output to the display 8 such as a navigation device for display. By outputting the optimum operation amount to the actuator 7, the host vehicle is automatically controlled to avoid a collision.

  The database unit 6 stores a relative distance, a relative speed, and a vehicle body width of the target detected by the millimeter wave radar 9 and the in-vehicle camera 10 mounted on the host vehicle in time series. Furthermore, various sensors 11 that monitor the running state of the host vehicle, for example, vehicle speed sensors, brake amount sensors, accelerator opening sensors, and GPS measured values are stored in time series. When the collision risk evaluation unit 4 determines that the risk of collision is high, the alarm 11 may be driven to give a warning to the driver.

  The database unit 6 can store not only the sensor output but also the value actually input to the actuator 7. As a result, it is possible to detect a deviation between the value actually input to the actuator 7 and the actual drive value detected by the sensor 11. This deviation can be used as a deviation value when creating the vehicle model.

  When the optimum operation amount search unit 5 detects the optimum operation amount of the actuator 7, it may be displayed on the display 8 to notify the driver. The driver controls the brake, accelerator, and steering to avoid the collision by referring to the displayed operation amount, or the detected optimum operation amount is forcibly input to the actuator 7 to automatically collide The avoidance may be determined by the driver's choice. In this case, for example, a “mode selection switch” may be provided in the vehicle so that the driver can select between collision avoidance by own operation or forced collision avoidance.

  FIG. 2 is a block diagram showing details of the optimum operation amount search unit 5 of FIG. The optimum operation amount search unit 5 of the present embodiment is calculated by an operation amount candidate value calculation unit 51 and an operation amount candidate value calculation unit 51 that calculate operation amount candidate values for brake, accelerator, and steering according to an optimization algorithm. The model prediction unit 52 that predicts the future position of the vehicle after a predetermined time by inputting the obtained value into the host vehicle model, and the vehicle future position and the target future position created by the model prediction unit 52 are input to the evaluation formula And an evaluation unit 53 for evaluating the risk of collision.

  The evaluation unit 53 feeds back an evaluation result to the manipulated variable candidate value calculation unit 51 in order to search for an evaluation value optimized based on the optimization algorithm. Any optimization algorithm can be selected, for example, the least squares method, genetic algorithm, etc. The selection of manipulated variable candidate values, evaluation value feedback methods, etc. are arbitrarily determined by the selected optimization algorithm. Is done.

  FIG. 3 is a flowchart for explaining the operation of the collision avoidance control apparatus shown in FIGS. 1 and 2. This series of processing is repeatedly performed at regular time intervals in a control device that manages and controls the running state of the vehicle. Or you may make it start when a target is detected ahead of the own vehicle by millimeter wave radar or a vehicle-mounted camera. Here, the former case will be described.

  When the process is started (step S1), the target future prediction unit 2 shown in FIG. 1 determines whether or not the target exists ahead of the host vehicle with reference to the database unit 6 (step S2). If the target does not exist, the process moves to step S9 as it is and ends the process. If the target exists (YES in step S2), the relative speed, relative distance, and target width of the target with respect to the host vehicle are detected (step S3). Since the values of the millimeter wave radar 9, the in-vehicle camera 10, and the various sensors 11 of the vehicle are always input and stored in the database unit 6, the target future prediction unit 2 refers to the database unit 6 to indicate the presence of the target. , And relative speed, relative distance, and target width can be detected.

  Next, the lateral movement distance of the target per predetermined time is calculated from the road curve R calculated from the steering angle of the host vehicle (step S4). This is caused by considering the movement of the target with respect to the position of the own vehicle at the coordinate 0 point, and the lateral movement distance per predetermined time of the own vehicle accompanying the curve R is represented by the lateral movement distance of the target. . The ground coordinates with the position of the own vehicle as the coordinate 0 point will be described later.

  In the next step S5, the future position of the target after a predetermined time set in advance is calculated based on the relative speed, relative distance, and lateral movement distance of the target. The target model is identified through steps S3 to S5.

  Next, with reference to the database unit 6, a movable range of the host vehicle after a predetermined time is set (step S6). FIG. 4 is a diagram illustrating a method for setting the movable range of the host vehicle. When the current vehicle position is indicated by A0, the position after a predetermined time when traveling at the current speed is indicated by A1, the position when the brake is maximized is A2, and the position when the accelerator is maximized is indicated. This is indicated by A3. The range B-B 'is obtained by setting the steering angle to the maximum left and right. In this case, the movable range of the own vehicle is an area indicated by oblique lines including the position A3.

  In step S7 of FIG. 3, whether or not the future position of the target exists within the movable range of the vehicle shown in FIG. 4, or how close the future position of the target is to the movable range, To evaluate. That is, the risk of collision is evaluated. In step S8, based on the evaluation value in step S7, it is determined whether or not a collision avoidance operation for the host vehicle is necessary. This evaluation is performed in the collision risk evaluation unit 4 shown in FIG. If it is determined that there is no need for the avoidance operation (NO in step S8), the process is terminated as it is (step S9).

  If it is determined in step S8 that a collision avoidance operation is necessary (YES in step S8), an optimum operation amount search process for avoiding a collision is executed in step S10. When the optimum operation amount is retrieved in step S10, the value is input to the operation target, that is, the own vehicle actuator in step S11, and the process is terminated (step S9). Alternatively, in step S11, the optimum operation amount may be displayed on the display.

  FIG. 5 is a flowchart showing details of the optimum operation amount search processing step S10 shown in FIG. First, in step S20, based on the selected optimization algorithm, candidate values that are temporarily input to the host vehicle model are extracted for the brake, accelerator, and steering. The process of step S20 is performed in the operation amount candidate value calculation unit 51 in FIG. In the next step S21, the future position after a predetermined time of the own vehicle is predicted by inputting the candidate value extracted in step S20 into the own vehicle model. The own vehicle model is identified from data relating to the traveling state of the own vehicle stored in the database unit 6. The process of step S21 is performed in the model prediction unit 52 in FIG.

  As described above, when the future position after a predetermined time of the host vehicle is predicted for the initial value of the operation amount, in the next step S22, the future position after the predetermined time of the target identified by the target model is determined. The relationship, i.e. the risk of collision, is evaluated. In this case, if the future positions of the two overlap or are close to each other, the risk evaluation becomes large. As the future positions of the two move away, the risk assessment becomes smaller.

  In step S23, it is determined whether or not the evaluation value obtained in step S22 is based on the first evaluation. If it is based on the first evaluation (YES in step S23), the process returns to step S20 and the subsequent steps are repeated. This is because it is not possible to determine whether or not the candidate value input to obtain the evaluation is the optimum value by only one evaluation in step S22. If NO in step S23, that is, if the previous evaluation value (N-1) exists for the current evaluation value N, in step S24, the previous and current evaluation values (N-1) and N are To be compared.

  If N <(N-1) in step S25 (YES in step S5), it is determined that the current candidate value is a value that can avoid danger more than the previous candidate value. In S26, it is determined whether or not the evaluation value N is sufficiently small to avoid danger, that is, whether or not the evaluation value N is smaller than a preset threshold value in the optimization algorithm. If it is determined that the value is smaller than the threshold value (YES in step S26), the candidate value of the operation amount used for obtaining the evaluation value N is output as the optimum operation amount (step S27). If N ≧ (N−1) in step S25, step S20 and subsequent steps are executed again in order to search for a more optimal candidate value. Similarly, if NO (N> threshold) in step S26, step S20 and subsequent steps are executed again.

  FIG. 6 is a diagram for explaining a method for evaluating the risk of collision. When the own vehicle and the other vehicle are in the relationship shown in the figure, when an operation amount that causes the own vehicle model 100M to move in the direction of the other vehicle (target) model 200M is input to the own vehicle model 100M, for example, steering is performed. The evaluation value becomes large because the risk of collision increases when turning to the right, and the evaluation value becomes small when the steering is turned to the left. Therefore, by searching for the operation amount that minimizes the evaluation value according to the optimization algorithm, the operation amount optimal for collision avoidance is calculated.

  FIG. 7 is a diagram for explaining a future position prediction method for the other vehicle (target) 200. By observing the other vehicle 200 mounted on the own vehicle 100 with, for example, a millimeter wave radar, the relative speed and distance of the other vehicle 200 with respect to the own vehicle 100 can be detected. By setting the relative speed thus detected as the speed of the other vehicle 200 and using the own vehicle 100 as the origin of coordinates, the future position of the other vehicle 200 ′ after a predetermined time can be predicted.

  FIG. 8 is a diagram for explaining selection of the optimum coordinate system for calculating the future position of the other vehicle. As shown in FIG. 8A, when the vehicle 100 is turning, if the coordinates are fixed to the vehicle 100, the future relative relationship between the vehicle and the other vehicle is different from the vehicle 100 ′. This is indicated by a car 200 '. In this case, in order to perform the coordinate calculation of the other vehicle 200 ′, the movement of the other vehicle 200 ′ needs to be expressed in the r−θ coordinate system. Furthermore, even if the other vehicle is traveling at a constant speed, it is necessary to correct the direction of movement of the other vehicle 200 by the rotation of the host vehicle 100, and the calculation becomes very complicated.

  Therefore, as shown in FIG. 8B, the future position of the vehicle is taken into consideration when calculating the future vehicle position 200 ′ by setting the current position of the host vehicle 100 as the coordinate 0 (reference) point. Since it is not necessary, the operation can be performed simply.

  FIG. 9 is a diagram for explaining the necessity of correction by rotation of the own vehicle in the calculation of the future other vehicle position based on the current own vehicle position shown in FIG. As described above, the position of the future vehicle can be calculated simply by using the current vehicle position as a reference. However, in the radar such as millimeter wave, the relative position and the relative speed of the target are output. There is no lateral speed output. Accordingly, it is necessary to calculate the lateral movement amount from the past relative position information, but at that time, it is necessary to perform correction by rotation of the own vehicle.

  For example, consider a road 300 with a wall 310 inside as shown in FIG. When the millimeter wave radar captures and outputs the wall surface 310 as a target, the inner wall surface 310 becomes a predicted course of collision with the host vehicle 100 as shown in FIG. Since the own vehicle 100 is rotating at this time, it is possible to prevent an erroneous collision determination due to a shift in the lateral position by performing the correction.

When calculating the future position of the target after a predetermined time in step S5 of FIG. 3, the future position to be obtained is the future position at the time when the evaluation value of the degree of risk is the highest, that is, the time when the possibility of collision is the highest. . Therefore, this time corresponds to the predetermined time in step S5. This time point is obtained by obtaining a time (pre-time) until the host vehicle and the target re-approach according to (distance / relative speed).
(Pre-time) = (distance between own vehicle and target) / relative speed (1)

  When the time (pre-time) is obtained according to the equation (1), the future position of the target at the time (pre-time) is then calculated. When there are a plurality of targets, (pre-time) is obtained for each target, and a future position after (pre-time) for each target is obtained. When the target is another vehicle, it is necessary to adjust the collision range of the other vehicle in consideration of the movement of the other vehicle until the collision in evaluating the risk of collision between the own vehicle and the other vehicle.

FIG. 10 shows a case where the time to the collision is short, that is, (pre-time) is small (FIG. 10A), and a case where the time to the collision is long, that is, (pre-time) is large (FIG. 10). (B)) shows collision ranges 400a and 400b calculated for other vehicles. The collision ranges 400a and 400b are calculated according to the following equation (2). In the equation (2), the conforming value A is a margin based on the width of the vehicle, and the conforming value B is a margin for adjusting the collision range. If the collision range is too large, false alarms frequently occur and the driver is confused. Therefore, the occurrence of the false alarms is adjusted by adjusting the fitness value B.
Collision range = conformity value A + conformance value B × (pre-time) (2)

The risk evaluation executed in step S7 in FIG. 3 is performed based on the evaluation value F calculated using (pre-time). A method for calculating the evaluation value F is shown below. The conforming value C is a constant set to display the evaluation value F in 4 digits. When the evaluation result F is displayed in 3 digits, the conforming value C set for 3 digits may be used.
Evaluation value F = Adapted value C / (pre-time) (3)

  FIG. 11 shows the magnitude of the evaluation value F based on the magnitude of (pre-time). Since the evaluation value F is inversely proportional to (pre-time) as apparent from the above equation (3), as shown in FIG. 11, when the other vehicle 200a is close to the own vehicle 100 (pre- (time: small), the evaluation value F increases. When the other vehicle 200b is far from the host vehicle 100 (pre-time: large), the evaluation value F decreases.

However, in the method of calculating the evaluation value F according to the above equation (3), the evaluation value is the same regardless of whether the host vehicle collides with the target from the front or gives up the target, which is problematic in practice. Therefore, this problem is solved by including the distance between the vehicle and the target in the evaluation value. The evaluation formula in this case is the following formula (4). The conforming value D shown below is a constant set to display the evaluation value F in 4 digits. When the evaluation value F is displayed in 3 digits, the conforming value D set for 3 digits is used.
Evaluation value F = Applicable value C / (pre-time) + Applicable value D × (distance between own vehicle and target) ² (4)

  For example, as shown in FIG. 12, when the collision range 400 is large and movement into the collision range is unavoidable due to the movement of the host vehicle 100, the evaluation value F is calculated according to the above equation (4), It is possible to calculate the route 500 that gives up the collision range.

  The collision risk evaluation unit 4 shown in FIG. 1 can calculate an evaluation value F in a state where the current brake, accelerator, and steering amount of the host vehicle are maintained by using the above-described evaluation expressions (3) and (4). is there. Therefore, the alarm 11 is driven based on the value of the evaluation value F, and an appropriate alarm output such as an alarm sound such as a seat belt drive, a hazard output, a brake lamp point, or the like can be performed.

  FIG. 13 illustrates the relationship between the evaluation value F and the alarm output. As an example of alarm display, “Automatic brake is active” at warning level 3 and “There is a danger of collision. Reduce the speed” at warning level 2 are displayed by voice or on the voice and display. Display text. Further, at the display warning level 1, it is possible to make only the sound effect such as “puff” and not display on the display.

  The calculation formulas (3) and (4) for the evaluation value F have been described on the premise that they are used in the collision risk evaluation unit 4 in FIGS. 11 and 12, but these evaluation formulas are optimal operations. Similarly, the evaluation unit 53 (see FIG. 2) of the quantity search unit 5 (see FIG. 1) can be used for calculating the evaluation value of the risk of collision.

  FIG. 14 shows a processing procedure for obtaining the evaluation value F for each target when there are a plurality of targets for the own vehicle. First, in step P1, the evaluation value F is calculated for the target 1. In step P1, the calculation step P11 of (pre-time [1]) for the target 1, the position prediction step P12 of the host vehicle and the target 1 after (pre-time [1]) seconds, and the evaluation value calculation for the target 1 are calculated. Step P13 is included.

  When the evaluation value for the target 1 is calculated in this way, in the next step P2, the same processing as in steps P11 to P13 is performed on the target 2 to calculate the evaluation formula for the target 2. Similarly, in step P3, an evaluation value for the target 3 is calculated. The above processing is repeated for all targets captured by the millimeter wave radar (step P4), and the evaluation value calculation processing for a plurality of targets is completed.

The block diagram which shows the structure of the collision avoidance control apparatus concerning one Embodiment of this invention. The block diagram which shows the detail of the optimal operation amount search part shown in FIG. The flowchart for operation | movement description in the collision avoidance control apparatus of FIG. The figure which shows the example of a setting of the movable range of the own vehicle. The flowchart which shows the detail of the search process of the optimal operation amount by model prediction control. The figure for demonstrating the evaluation method of a collision risk. The figure for demonstrating the future position prediction method. The figure for demonstrating the selection method of a coordinate system. The figure for demonstrating the correction | amendment accompanying rotation of the own vehicle. The figure for demonstrating the collision range calculation of another vehicle. The figure which shows the relationship between pre-time and an evaluation value. The figure for demonstrating the calculation method of the path | route when a collision range is large. The figure which shows an alarm output example. The flowchart which shows the evaluation value calculation procedure about multiple targets.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Collision avoidance control apparatus 2 Target future prediction part 3 Own vehicle future prediction part 4 Collision risk evaluation part 5 Optimal operation amount search part 6 Database part 7 Actuator 8 Display 9 Millimeter wave radar 10 Car-mounted camera 11 Sensor 51 Operation amount candidate value calculation Part 52 model prediction part 53 evaluation part

Claims (5)

A database unit for storing data relating to the traveling state of the host vehicle and data relating to the distance between the target and the host vehicle and the relative speed of the target;
A target future prediction unit that predicts the future position of the target based on the data stored in the database unit;
A vehicle future prediction unit that predicts a future position of the vehicle based on data stored in the database unit;
A collision risk evaluation unit that evaluates the risk of collision between the vehicle and the target based on the output of the future prediction unit of the target and the vehicle;
A collision avoidance control device for a vehicle, comprising: an optimal operation amount search unit that performs model predictive control based on the output of the collision risk evaluation unit and searches for an optimal operation amount related to an actuator of the own vehicle for collision avoidance.   The collision avoidance control device according to claim 1, wherein the optimum operation amount search unit includes an operation amount candidate value calculation unit that calculates an operation amount candidate value based on an arbitrarily selected optimization algorithm, and a predetermined value of the host vehicle. A model prediction unit that predicts the future position after time; and an evaluation unit that evaluates the risk of collision between the predicted future position of the vehicle and the future position of the other vehicle, and based on the optimization algorithm, the evaluation A collision avoidance control apparatus for a vehicle, wherein an operation amount with the lowest risk of collision is searched by feeding back a partial output to the operation amount candidate calculation unit.   2. The collision avoidance control device according to claim 1, wherein the optimum operation amount search unit searches for an operation amount related to a brake and a steering as an optimum operation amount of an actuator for avoiding a collision between the host vehicle and a target. A collision avoidance control device for a vehicle.   The collision avoidance control device for a vehicle according to claim 1, wherein the collision risk evaluation unit evaluates a collision risk based on a time required for the vehicle and the target to be closest to each other. Vehicle collision avoidance control device.   5. The collision avoidance control apparatus for a vehicle according to claim 4, wherein when there are a plurality of targets, the time is calculated for each of the own vehicle and the plurality of targets.

Images